Skip to content

Latest commit

 

History

History
48 lines (41 loc) · 2.78 KB

README.md

File metadata and controls

48 lines (41 loc) · 2.78 KB

LSENeRF-data-formatter

This repo contains the code to format the data in the paper LSENeRF

Paper | Webpage | Method code

Download the dataset

For details of the downloaded files and formatted data see here

Mandatory files

Scenes

Outdoors Indoors
Bag Dragon Max
Courtyard Grad Lounge
Bicycle Lab
Engineer Building Presentation Room
House Teddy Grass

Preliminaries

Make an environment with the packages in requirements.txt installed

Formating LSENeRF dataset

To format a scene in our dataset:

  1. Download ecam_intrinsics.json, rel_extrinsics and a scene from the above
  2. Fill out the fields in format_raw.sh, the fields details are provided in the comments of the shell script
  3. run the below
bash format_raw.sh

Formatting EVIMOv2 dataset

  1. Download the flea3_7 and samsung_mono NPZ datatset here
  2. Download NPZ extrinsics overlay here. This is available on the same page as 1.
    1. The NPZ extrinsics overlay is camera transform for camera -> rig, the rig is equivalent to world
  3. Extract NPZ extrinsics overlay from 2. on top of the NPZ dataset so that dataset_extrinsics.npz is available in the dataset scene folders
  4. Fill the neccessary fields in format_evimo.sh
  5. Format a scene in the evimovs dataset with
bash format_evimo.sh

Note that only these EVIMOv2 scenes are supported (see the evimov2_dataset directory):

  • depth_var_1_lr_000000
  • scene7_00_000001
  • scene8_01_000000

For details of the formatted scenes, see here